Embed Size (px)
Citation preview
Whole-Exome Sequencing Technical Details
Jim Mullikin Director NIH Intramural Sequencing Center
Head Comparative Genomics Unit
Whole Exome Sequencing Why
bull Focuses on the part of the genome we understand best the exons of genes
bull Exomes are ideal to help us understand high-penetrance allelic variation and its relationship to phenotype
bull A whole exome is 16 the cost of whole genome and 115 the amount of data
Biesecker et al Genome Biology 2011 12128
1
Twinbrook ResearchBuilding
5625FishersLaneRockvilleMD
NISCoccupies entire5th floor
NISCSequenceProduction
Feb 2010
2
Sequencing applications processed through NGS production pipeline from April 2009 to August 2011
Number of Samples
Total FY11 Sequenced 169 billion reads
Total NISC Lifetime Sequenced 255 billion reads
4
CimarronSoftwarebasedNextGen LIMS
Computational Resources for 6 GAiiX and 3 HiSeq2000
bull Linux cluster ndash 1000 cores
bull 250 for production ndash 900TB disk
bull 250TB for production with 75TB available bull 15TBmonth long term storage
ndash Network bull 1 and 10 Gigabit-Ethernet
5
Number of Samples
Sequencing applications processed through NGS production pipeline from April 2009 to August 2011
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
6
Library Preparation
httpwwwilluminacomsupportliteratureilmn
AgilentTechnologies SureSelect method
WholeͲexome kit 38Mband50Mb
httpcpliteratureagilentcomlitwebpdf5990-3532ENpdf
7
bull Illumina TruSeq Exome Enrichment bull 62Mb of exome targeted
httpwwwilluminacomsupportliteratureilmn
NISC Exome Process Tag 1
Indexed Libraries Tag 2 Tag 3 Tag 4 Tag 5 Tag 6 Balance
and Pool
Illumina TruSeq Exome Enrichment
Sequence on two lanes
8
Cluster Generation
httpwwwilluminacomsupportliteratureilmn
Sequencing
httpwwwilluminacomsupportliteratureilmn
9
HiSeq Flow Cell
bull 8 lanes bull 15 billion clusters bull Up to 300Gb per flow cell
Sequencing and Data Processing ~150M clusters per lane
2x100 base paired-end reads and the index tag
Tag 1 Tag 2 Tag 3 Tag 4 Tag 5 Tag 6
Demultiplex
Alignment ELAND to Human Genome
Total sequence per sample ~10Gb from two lanes Over 100X coverage of targeted regions
10
Read Depth-of-Coverage
400
ITGA6
Read Depth-of-Coverage
400
ITGA6
11
Targeted Regions Depth-of-Coverage Histogram
0
0005
001
0015
002
0 50 100 150 200
Depth of Coverage
Frac
tion
per B
in
TruSeq_example Poisson
Refining the Alignment (diagCM)
bull ELAND is part of the standard pipeline
bull ELAND accurately places reads in the correct genomic location
bull Use cross-match a Smith-Waterman aligner to improve local alignment
12
Eland version XXX
ELAND Aligned Reads
eland_ms_32 Casava 17
Cross-match Improved Alignment
13
Comparison of Aligners (simulated exome data)
Perc
enta
ge C
orre
ct
85
90
9510
0
-6 -4 -2 0 2 4 6
Variant Size
BayesianGenotypeCalling MostProbableGenotype
ACCTTCCTCGAGGCTAGCCTAGCCGGGGTGGGCAAGGCCTTCCGGGGGGGTAACCTCCTCCAAAACCTCCCTATGCCCCCCCAATTTTTTATATATATCCTCCTCCAA A
A
A A
A
A
A A
A A
A A
P(Genotype|Data)
AT 0 AA -14
Difference between most probable AA 10-6
10-15 and next most probable genotype AC 10-15AG AC -34
AT 099999981 AG -34 10-75CC TT -124 MPG Score = 14 10-75 CT -138CG 10-60CT Convert to natural Log GT -138
GG 10-75 Probabilities and Sort CC -172 10-60GT CG -172 10-54TT GG -172
14
MPG of Haploid Regions
bull Human autosomes are normally diploid bull MPG is designed to call two alleles for the
autosomes and X chromosome if the sample is from a female
bull For samples from males MPG is run in haploid mode on the non-sudoautosomalregions of X and Y
bull Thus only testing for the four nucleotides
Exome Coverage versus Input Sequence
15
Num
ber o
f Sam
ples
SureSelect 38Mb
SureSelect 50Mb
Illumina TruSeq 62Mb
0 20
0 40
0 60
0 80
0 10
00
Whole Exomes Processed at NISC
Mar 2010 Dec 2010 Sep 2011
NBEAL2
16
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
Twinbrook ResearchBuilding
5625FishersLaneRockvilleMD
NISCoccupies entire5th floor
NISCSequenceProduction
Feb 2010
2
Sequencing applications processed through NGS production pipeline from April 2009 to August 2011
Number of Samples
Total FY11 Sequenced 169 billion reads
Total NISC Lifetime Sequenced 255 billion reads
4
CimarronSoftwarebasedNextGen LIMS
Computational Resources for 6 GAiiX and 3 HiSeq2000
bull Linux cluster ndash 1000 cores
bull 250 for production ndash 900TB disk
bull 250TB for production with 75TB available bull 15TBmonth long term storage
ndash Network bull 1 and 10 Gigabit-Ethernet
5
Number of Samples
Sequencing applications processed through NGS production pipeline from April 2009 to August 2011
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
6
Library Preparation
httpwwwilluminacomsupportliteratureilmn
AgilentTechnologies SureSelect method
WholeͲexome kit 38Mband50Mb
httpcpliteratureagilentcomlitwebpdf5990-3532ENpdf
7
bull Illumina TruSeq Exome Enrichment bull 62Mb of exome targeted
httpwwwilluminacomsupportliteratureilmn
NISC Exome Process Tag 1
Indexed Libraries Tag 2 Tag 3 Tag 4 Tag 5 Tag 6 Balance
and Pool
Illumina TruSeq Exome Enrichment
Sequence on two lanes
8
Cluster Generation
httpwwwilluminacomsupportliteratureilmn
Sequencing
httpwwwilluminacomsupportliteratureilmn
9
HiSeq Flow Cell
bull 8 lanes bull 15 billion clusters bull Up to 300Gb per flow cell
Sequencing and Data Processing ~150M clusters per lane
2x100 base paired-end reads and the index tag
Tag 1 Tag 2 Tag 3 Tag 4 Tag 5 Tag 6
Demultiplex
Alignment ELAND to Human Genome
Total sequence per sample ~10Gb from two lanes Over 100X coverage of targeted regions
10
Read Depth-of-Coverage
400
ITGA6
Read Depth-of-Coverage
400
ITGA6
11
Targeted Regions Depth-of-Coverage Histogram
0
0005
001
0015
002
0 50 100 150 200
Depth of Coverage
Frac
tion
per B
in
TruSeq_example Poisson
Refining the Alignment (diagCM)
bull ELAND is part of the standard pipeline
bull ELAND accurately places reads in the correct genomic location
bull Use cross-match a Smith-Waterman aligner to improve local alignment
12
Eland version XXX
ELAND Aligned Reads
eland_ms_32 Casava 17
Cross-match Improved Alignment
13
Comparison of Aligners (simulated exome data)
Perc
enta
ge C
orre
ct
85
90
9510
0
-6 -4 -2 0 2 4 6
Variant Size
BayesianGenotypeCalling MostProbableGenotype
ACCTTCCTCGAGGCTAGCCTAGCCGGGGTGGGCAAGGCCTTCCGGGGGGGTAACCTCCTCCAAAACCTCCCTATGCCCCCCCAATTTTTTATATATATCCTCCTCCAA A
A
A A
A
A
A A
A A
A A
P(Genotype|Data)
AT 0 AA -14
Difference between most probable AA 10-6
10-15 and next most probable genotype AC 10-15AG AC -34
AT 099999981 AG -34 10-75CC TT -124 MPG Score = 14 10-75 CT -138CG 10-60CT Convert to natural Log GT -138
GG 10-75 Probabilities and Sort CC -172 10-60GT CG -172 10-54TT GG -172
14
MPG of Haploid Regions
bull Human autosomes are normally diploid bull MPG is designed to call two alleles for the
autosomes and X chromosome if the sample is from a female
bull For samples from males MPG is run in haploid mode on the non-sudoautosomalregions of X and Y
bull Thus only testing for the four nucleotides
Exome Coverage versus Input Sequence
15
Num
ber o
f Sam
ples
SureSelect 38Mb
SureSelect 50Mb
Illumina TruSeq 62Mb
0 20
0 40
0 60
0 80
0 10
00
Whole Exomes Processed at NISC
Mar 2010 Dec 2010 Sep 2011
NBEAL2
16
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
Sequencing applications processed through NGS production pipeline from April 2009 to August 2011
Number of Samples
Total FY11 Sequenced 169 billion reads
Total NISC Lifetime Sequenced 255 billion reads
4
CimarronSoftwarebasedNextGen LIMS
Computational Resources for 6 GAiiX and 3 HiSeq2000
bull Linux cluster ndash 1000 cores
bull 250 for production ndash 900TB disk
bull 250TB for production with 75TB available bull 15TBmonth long term storage
ndash Network bull 1 and 10 Gigabit-Ethernet
5
Number of Samples
Sequencing applications processed through NGS production pipeline from April 2009 to August 2011
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
6
Library Preparation
httpwwwilluminacomsupportliteratureilmn
AgilentTechnologies SureSelect method
WholeͲexome kit 38Mband50Mb
httpcpliteratureagilentcomlitwebpdf5990-3532ENpdf
7
bull Illumina TruSeq Exome Enrichment bull 62Mb of exome targeted
httpwwwilluminacomsupportliteratureilmn
NISC Exome Process Tag 1
Indexed Libraries Tag 2 Tag 3 Tag 4 Tag 5 Tag 6 Balance
and Pool
Illumina TruSeq Exome Enrichment
Sequence on two lanes
8
Cluster Generation
httpwwwilluminacomsupportliteratureilmn
Sequencing
httpwwwilluminacomsupportliteratureilmn
9
HiSeq Flow Cell
bull 8 lanes bull 15 billion clusters bull Up to 300Gb per flow cell
Sequencing and Data Processing ~150M clusters per lane
2x100 base paired-end reads and the index tag
Tag 1 Tag 2 Tag 3 Tag 4 Tag 5 Tag 6
Demultiplex
Alignment ELAND to Human Genome
Total sequence per sample ~10Gb from two lanes Over 100X coverage of targeted regions
10
Read Depth-of-Coverage
400
ITGA6
Read Depth-of-Coverage
400
ITGA6
11
Targeted Regions Depth-of-Coverage Histogram
0
0005
001
0015
002
0 50 100 150 200
Depth of Coverage
Frac
tion
per B
in
TruSeq_example Poisson
Refining the Alignment (diagCM)
bull ELAND is part of the standard pipeline
bull ELAND accurately places reads in the correct genomic location
bull Use cross-match a Smith-Waterman aligner to improve local alignment
12
Eland version XXX
ELAND Aligned Reads
eland_ms_32 Casava 17
Cross-match Improved Alignment
13
Comparison of Aligners (simulated exome data)
Perc
enta
ge C
orre
ct
85
90
9510
0
-6 -4 -2 0 2 4 6
Variant Size
BayesianGenotypeCalling MostProbableGenotype
ACCTTCCTCGAGGCTAGCCTAGCCGGGGTGGGCAAGGCCTTCCGGGGGGGTAACCTCCTCCAAAACCTCCCTATGCCCCCCCAATTTTTTATATATATCCTCCTCCAA A
A
A A
A
A
A A
A A
A A
P(Genotype|Data)
AT 0 AA -14
Difference between most probable AA 10-6
10-15 and next most probable genotype AC 10-15AG AC -34
AT 099999981 AG -34 10-75CC TT -124 MPG Score = 14 10-75 CT -138CG 10-60CT Convert to natural Log GT -138
GG 10-75 Probabilities and Sort CC -172 10-60GT CG -172 10-54TT GG -172
14
MPG of Haploid Regions
bull Human autosomes are normally diploid bull MPG is designed to call two alleles for the
autosomes and X chromosome if the sample is from a female
bull For samples from males MPG is run in haploid mode on the non-sudoautosomalregions of X and Y
bull Thus only testing for the four nucleotides
Exome Coverage versus Input Sequence
15
Num
ber o
f Sam
ples
SureSelect 38Mb
SureSelect 50Mb
Illumina TruSeq 62Mb
0 20
0 40
0 60
0 80
0 10
00
Whole Exomes Processed at NISC
Mar 2010 Dec 2010 Sep 2011
NBEAL2
16
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
CimarronSoftwarebasedNextGen LIMS
Computational Resources for 6 GAiiX and 3 HiSeq2000
bull Linux cluster ndash 1000 cores
bull 250 for production ndash 900TB disk
bull 250TB for production with 75TB available bull 15TBmonth long term storage
ndash Network bull 1 and 10 Gigabit-Ethernet
5
Number of Samples
Sequencing applications processed through NGS production pipeline from April 2009 to August 2011
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
6
Library Preparation
httpwwwilluminacomsupportliteratureilmn
AgilentTechnologies SureSelect method
WholeͲexome kit 38Mband50Mb
httpcpliteratureagilentcomlitwebpdf5990-3532ENpdf
7
bull Illumina TruSeq Exome Enrichment bull 62Mb of exome targeted
httpwwwilluminacomsupportliteratureilmn
NISC Exome Process Tag 1
Indexed Libraries Tag 2 Tag 3 Tag 4 Tag 5 Tag 6 Balance
and Pool
Illumina TruSeq Exome Enrichment
Sequence on two lanes
8
Cluster Generation
httpwwwilluminacomsupportliteratureilmn
Sequencing
httpwwwilluminacomsupportliteratureilmn
9
HiSeq Flow Cell
bull 8 lanes bull 15 billion clusters bull Up to 300Gb per flow cell
Sequencing and Data Processing ~150M clusters per lane
2x100 base paired-end reads and the index tag
Tag 1 Tag 2 Tag 3 Tag 4 Tag 5 Tag 6
Demultiplex
Alignment ELAND to Human Genome
Total sequence per sample ~10Gb from two lanes Over 100X coverage of targeted regions
10
Read Depth-of-Coverage
400
ITGA6
Read Depth-of-Coverage
400
ITGA6
11
Targeted Regions Depth-of-Coverage Histogram
0
0005
001
0015
002
0 50 100 150 200
Depth of Coverage
Frac
tion
per B
in
TruSeq_example Poisson
Refining the Alignment (diagCM)
bull ELAND is part of the standard pipeline
bull ELAND accurately places reads in the correct genomic location
bull Use cross-match a Smith-Waterman aligner to improve local alignment
12
Eland version XXX
ELAND Aligned Reads
eland_ms_32 Casava 17
Cross-match Improved Alignment
13
Comparison of Aligners (simulated exome data)
Perc
enta
ge C
orre
ct
85
90
9510
0
-6 -4 -2 0 2 4 6
Variant Size
BayesianGenotypeCalling MostProbableGenotype
ACCTTCCTCGAGGCTAGCCTAGCCGGGGTGGGCAAGGCCTTCCGGGGGGGTAACCTCCTCCAAAACCTCCCTATGCCCCCCCAATTTTTTATATATATCCTCCTCCAA A
A
A A
A
A
A A
A A
A A
P(Genotype|Data)
AT 0 AA -14
Difference between most probable AA 10-6
10-15 and next most probable genotype AC 10-15AG AC -34
AT 099999981 AG -34 10-75CC TT -124 MPG Score = 14 10-75 CT -138CG 10-60CT Convert to natural Log GT -138
GG 10-75 Probabilities and Sort CC -172 10-60GT CG -172 10-54TT GG -172
14
MPG of Haploid Regions
bull Human autosomes are normally diploid bull MPG is designed to call two alleles for the
autosomes and X chromosome if the sample is from a female
bull For samples from males MPG is run in haploid mode on the non-sudoautosomalregions of X and Y
bull Thus only testing for the four nucleotides
Exome Coverage versus Input Sequence
15
Num
ber o
f Sam
ples
SureSelect 38Mb
SureSelect 50Mb
Illumina TruSeq 62Mb
0 20
0 40
0 60
0 80
0 10
00
Whole Exomes Processed at NISC
Mar 2010 Dec 2010 Sep 2011
NBEAL2
16
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
Number of Samples
Sequencing applications processed through NGS production pipeline from April 2009 to August 2011
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
6
Library Preparation
httpwwwilluminacomsupportliteratureilmn
AgilentTechnologies SureSelect method
WholeͲexome kit 38Mband50Mb
httpcpliteratureagilentcomlitwebpdf5990-3532ENpdf
7
bull Illumina TruSeq Exome Enrichment bull 62Mb of exome targeted
httpwwwilluminacomsupportliteratureilmn
NISC Exome Process Tag 1
Indexed Libraries Tag 2 Tag 3 Tag 4 Tag 5 Tag 6 Balance
and Pool
Illumina TruSeq Exome Enrichment
Sequence on two lanes
8
Cluster Generation
httpwwwilluminacomsupportliteratureilmn
Sequencing
httpwwwilluminacomsupportliteratureilmn
9
HiSeq Flow Cell
bull 8 lanes bull 15 billion clusters bull Up to 300Gb per flow cell
Sequencing and Data Processing ~150M clusters per lane
2x100 base paired-end reads and the index tag
Tag 1 Tag 2 Tag 3 Tag 4 Tag 5 Tag 6
Demultiplex
Alignment ELAND to Human Genome
Total sequence per sample ~10Gb from two lanes Over 100X coverage of targeted regions
10
Read Depth-of-Coverage
400
ITGA6
Read Depth-of-Coverage
400
ITGA6
11
Targeted Regions Depth-of-Coverage Histogram
0
0005
001
0015
002
0 50 100 150 200
Depth of Coverage
Frac
tion
per B
in
TruSeq_example Poisson
Refining the Alignment (diagCM)
bull ELAND is part of the standard pipeline
bull ELAND accurately places reads in the correct genomic location
bull Use cross-match a Smith-Waterman aligner to improve local alignment
12
Eland version XXX
ELAND Aligned Reads
eland_ms_32 Casava 17
Cross-match Improved Alignment
13
Comparison of Aligners (simulated exome data)
Perc
enta
ge C
orre
ct
85
90
9510
0
-6 -4 -2 0 2 4 6
Variant Size
BayesianGenotypeCalling MostProbableGenotype
ACCTTCCTCGAGGCTAGCCTAGCCGGGGTGGGCAAGGCCTTCCGGGGGGGTAACCTCCTCCAAAACCTCCCTATGCCCCCCCAATTTTTTATATATATCCTCCTCCAA A
A
A A
A
A
A A
A A
A A
P(Genotype|Data)
AT 0 AA -14
Difference between most probable AA 10-6
10-15 and next most probable genotype AC 10-15AG AC -34
AT 099999981 AG -34 10-75CC TT -124 MPG Score = 14 10-75 CT -138CG 10-60CT Convert to natural Log GT -138
GG 10-75 Probabilities and Sort CC -172 10-60GT CG -172 10-54TT GG -172
14
MPG of Haploid Regions
bull Human autosomes are normally diploid bull MPG is designed to call two alleles for the
autosomes and X chromosome if the sample is from a female
bull For samples from males MPG is run in haploid mode on the non-sudoautosomalregions of X and Y
bull Thus only testing for the four nucleotides
Exome Coverage versus Input Sequence
15
Num
ber o
f Sam
ples
SureSelect 38Mb
SureSelect 50Mb
Illumina TruSeq 62Mb
0 20
0 40
0 60
0 80
0 10
00
Whole Exomes Processed at NISC
Mar 2010 Dec 2010 Sep 2011
NBEAL2
16
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
Library Preparation
httpwwwilluminacomsupportliteratureilmn
AgilentTechnologies SureSelect method
WholeͲexome kit 38Mband50Mb
httpcpliteratureagilentcomlitwebpdf5990-3532ENpdf
7
bull Illumina TruSeq Exome Enrichment bull 62Mb of exome targeted
httpwwwilluminacomsupportliteratureilmn
NISC Exome Process Tag 1
Indexed Libraries Tag 2 Tag 3 Tag 4 Tag 5 Tag 6 Balance
and Pool
Illumina TruSeq Exome Enrichment
Sequence on two lanes
8
Cluster Generation
httpwwwilluminacomsupportliteratureilmn
Sequencing
httpwwwilluminacomsupportliteratureilmn
9
HiSeq Flow Cell
bull 8 lanes bull 15 billion clusters bull Up to 300Gb per flow cell
Sequencing and Data Processing ~150M clusters per lane
2x100 base paired-end reads and the index tag
Tag 1 Tag 2 Tag 3 Tag 4 Tag 5 Tag 6
Demultiplex
Alignment ELAND to Human Genome
Total sequence per sample ~10Gb from two lanes Over 100X coverage of targeted regions
10
Read Depth-of-Coverage
400
ITGA6
Read Depth-of-Coverage
400
ITGA6
11
Targeted Regions Depth-of-Coverage Histogram
0
0005
001
0015
002
0 50 100 150 200
Depth of Coverage
Frac
tion
per B
in
TruSeq_example Poisson
Refining the Alignment (diagCM)
bull ELAND is part of the standard pipeline
bull ELAND accurately places reads in the correct genomic location
bull Use cross-match a Smith-Waterman aligner to improve local alignment
12
Eland version XXX
ELAND Aligned Reads
eland_ms_32 Casava 17
Cross-match Improved Alignment
13
Comparison of Aligners (simulated exome data)
Perc
enta
ge C
orre
ct
85
90
9510
0
-6 -4 -2 0 2 4 6
Variant Size
BayesianGenotypeCalling MostProbableGenotype
ACCTTCCTCGAGGCTAGCCTAGCCGGGGTGGGCAAGGCCTTCCGGGGGGGTAACCTCCTCCAAAACCTCCCTATGCCCCCCCAATTTTTTATATATATCCTCCTCCAA A
A
A A
A
A
A A
A A
A A
P(Genotype|Data)
AT 0 AA -14
Difference between most probable AA 10-6
10-15 and next most probable genotype AC 10-15AG AC -34
AT 099999981 AG -34 10-75CC TT -124 MPG Score = 14 10-75 CT -138CG 10-60CT Convert to natural Log GT -138
GG 10-75 Probabilities and Sort CC -172 10-60GT CG -172 10-54TT GG -172
14
MPG of Haploid Regions
bull Human autosomes are normally diploid bull MPG is designed to call two alleles for the
autosomes and X chromosome if the sample is from a female
bull For samples from males MPG is run in haploid mode on the non-sudoautosomalregions of X and Y
bull Thus only testing for the four nucleotides
Exome Coverage versus Input Sequence
15
Num
ber o
f Sam
ples
SureSelect 38Mb
SureSelect 50Mb
Illumina TruSeq 62Mb
0 20
0 40
0 60
0 80
0 10
00
Whole Exomes Processed at NISC
Mar 2010 Dec 2010 Sep 2011
NBEAL2
16
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
bull Illumina TruSeq Exome Enrichment bull 62Mb of exome targeted
httpwwwilluminacomsupportliteratureilmn
NISC Exome Process Tag 1
Indexed Libraries Tag 2 Tag 3 Tag 4 Tag 5 Tag 6 Balance
and Pool
Illumina TruSeq Exome Enrichment
Sequence on two lanes
8
Cluster Generation
httpwwwilluminacomsupportliteratureilmn
Sequencing
httpwwwilluminacomsupportliteratureilmn
9
HiSeq Flow Cell
bull 8 lanes bull 15 billion clusters bull Up to 300Gb per flow cell
Sequencing and Data Processing ~150M clusters per lane
2x100 base paired-end reads and the index tag
Tag 1 Tag 2 Tag 3 Tag 4 Tag 5 Tag 6
Demultiplex
Alignment ELAND to Human Genome
Total sequence per sample ~10Gb from two lanes Over 100X coverage of targeted regions
10
Read Depth-of-Coverage
400
ITGA6
Read Depth-of-Coverage
400
ITGA6
11
Targeted Regions Depth-of-Coverage Histogram
0
0005
001
0015
002
0 50 100 150 200
Depth of Coverage
Frac
tion
per B
in
TruSeq_example Poisson
Refining the Alignment (diagCM)
bull ELAND is part of the standard pipeline
bull ELAND accurately places reads in the correct genomic location
bull Use cross-match a Smith-Waterman aligner to improve local alignment
12
Eland version XXX
ELAND Aligned Reads
eland_ms_32 Casava 17
Cross-match Improved Alignment
13
Comparison of Aligners (simulated exome data)
Perc
enta
ge C
orre
ct
85
90
9510
0
-6 -4 -2 0 2 4 6
Variant Size
BayesianGenotypeCalling MostProbableGenotype
ACCTTCCTCGAGGCTAGCCTAGCCGGGGTGGGCAAGGCCTTCCGGGGGGGTAACCTCCTCCAAAACCTCCCTATGCCCCCCCAATTTTTTATATATATCCTCCTCCAA A
A
A A
A
A
A A
A A
A A
P(Genotype|Data)
AT 0 AA -14
Difference between most probable AA 10-6
10-15 and next most probable genotype AC 10-15AG AC -34
AT 099999981 AG -34 10-75CC TT -124 MPG Score = 14 10-75 CT -138CG 10-60CT Convert to natural Log GT -138
GG 10-75 Probabilities and Sort CC -172 10-60GT CG -172 10-54TT GG -172
14
MPG of Haploid Regions
bull Human autosomes are normally diploid bull MPG is designed to call two alleles for the
autosomes and X chromosome if the sample is from a female
bull For samples from males MPG is run in haploid mode on the non-sudoautosomalregions of X and Y
bull Thus only testing for the four nucleotides
Exome Coverage versus Input Sequence
15
Num
ber o
f Sam
ples
SureSelect 38Mb
SureSelect 50Mb
Illumina TruSeq 62Mb
0 20
0 40
0 60
0 80
0 10
00
Whole Exomes Processed at NISC
Mar 2010 Dec 2010 Sep 2011
NBEAL2
16
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
Cluster Generation
httpwwwilluminacomsupportliteratureilmn
Sequencing
httpwwwilluminacomsupportliteratureilmn
9
HiSeq Flow Cell
bull 8 lanes bull 15 billion clusters bull Up to 300Gb per flow cell
Sequencing and Data Processing ~150M clusters per lane
2x100 base paired-end reads and the index tag
Tag 1 Tag 2 Tag 3 Tag 4 Tag 5 Tag 6
Demultiplex
Alignment ELAND to Human Genome
Total sequence per sample ~10Gb from two lanes Over 100X coverage of targeted regions
10
Read Depth-of-Coverage
400
ITGA6
Read Depth-of-Coverage
400
ITGA6
11
Targeted Regions Depth-of-Coverage Histogram
0
0005
001
0015
002
0 50 100 150 200
Depth of Coverage
Frac
tion
per B
in
TruSeq_example Poisson
Refining the Alignment (diagCM)
bull ELAND is part of the standard pipeline
bull ELAND accurately places reads in the correct genomic location
bull Use cross-match a Smith-Waterman aligner to improve local alignment
12
Eland version XXX
ELAND Aligned Reads
eland_ms_32 Casava 17
Cross-match Improved Alignment
13
Comparison of Aligners (simulated exome data)
Perc
enta
ge C
orre
ct
85
90
9510
0
-6 -4 -2 0 2 4 6
Variant Size
BayesianGenotypeCalling MostProbableGenotype
ACCTTCCTCGAGGCTAGCCTAGCCGGGGTGGGCAAGGCCTTCCGGGGGGGTAACCTCCTCCAAAACCTCCCTATGCCCCCCCAATTTTTTATATATATCCTCCTCCAA A
A
A A
A
A
A A
A A
A A
P(Genotype|Data)
AT 0 AA -14
Difference between most probable AA 10-6
10-15 and next most probable genotype AC 10-15AG AC -34
AT 099999981 AG -34 10-75CC TT -124 MPG Score = 14 10-75 CT -138CG 10-60CT Convert to natural Log GT -138
GG 10-75 Probabilities and Sort CC -172 10-60GT CG -172 10-54TT GG -172
14
MPG of Haploid Regions
bull Human autosomes are normally diploid bull MPG is designed to call two alleles for the
autosomes and X chromosome if the sample is from a female
bull For samples from males MPG is run in haploid mode on the non-sudoautosomalregions of X and Y
bull Thus only testing for the four nucleotides
Exome Coverage versus Input Sequence
15
Num
ber o
f Sam
ples
SureSelect 38Mb
SureSelect 50Mb
Illumina TruSeq 62Mb
0 20
0 40
0 60
0 80
0 10
00
Whole Exomes Processed at NISC
Mar 2010 Dec 2010 Sep 2011
NBEAL2
16
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
HiSeq Flow Cell
bull 8 lanes bull 15 billion clusters bull Up to 300Gb per flow cell
Sequencing and Data Processing ~150M clusters per lane
2x100 base paired-end reads and the index tag
Tag 1 Tag 2 Tag 3 Tag 4 Tag 5 Tag 6
Demultiplex
Alignment ELAND to Human Genome
Total sequence per sample ~10Gb from two lanes Over 100X coverage of targeted regions
10
Read Depth-of-Coverage
400
ITGA6
Read Depth-of-Coverage
400
ITGA6
11
Targeted Regions Depth-of-Coverage Histogram
0
0005
001
0015
002
0 50 100 150 200
Depth of Coverage
Frac
tion
per B
in
TruSeq_example Poisson
Refining the Alignment (diagCM)
bull ELAND is part of the standard pipeline
bull ELAND accurately places reads in the correct genomic location
bull Use cross-match a Smith-Waterman aligner to improve local alignment
12
Eland version XXX
ELAND Aligned Reads
eland_ms_32 Casava 17
Cross-match Improved Alignment
13
Comparison of Aligners (simulated exome data)
Perc
enta
ge C
orre
ct
85
90
9510
0
-6 -4 -2 0 2 4 6
Variant Size
BayesianGenotypeCalling MostProbableGenotype
ACCTTCCTCGAGGCTAGCCTAGCCGGGGTGGGCAAGGCCTTCCGGGGGGGTAACCTCCTCCAAAACCTCCCTATGCCCCCCCAATTTTTTATATATATCCTCCTCCAA A
A
A A
A
A
A A
A A
A A
P(Genotype|Data)
AT 0 AA -14
Difference between most probable AA 10-6
10-15 and next most probable genotype AC 10-15AG AC -34
AT 099999981 AG -34 10-75CC TT -124 MPG Score = 14 10-75 CT -138CG 10-60CT Convert to natural Log GT -138
GG 10-75 Probabilities and Sort CC -172 10-60GT CG -172 10-54TT GG -172
14
MPG of Haploid Regions
bull Human autosomes are normally diploid bull MPG is designed to call two alleles for the
autosomes and X chromosome if the sample is from a female
bull For samples from males MPG is run in haploid mode on the non-sudoautosomalregions of X and Y
bull Thus only testing for the four nucleotides
Exome Coverage versus Input Sequence
15
Num
ber o
f Sam
ples
SureSelect 38Mb
SureSelect 50Mb
Illumina TruSeq 62Mb
0 20
0 40
0 60
0 80
0 10
00
Whole Exomes Processed at NISC
Mar 2010 Dec 2010 Sep 2011
NBEAL2
16
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
Read Depth-of-Coverage
400
ITGA6
Read Depth-of-Coverage
400
ITGA6
11
Targeted Regions Depth-of-Coverage Histogram
0
0005
001
0015
002
0 50 100 150 200
Depth of Coverage
Frac
tion
per B
in
TruSeq_example Poisson
Refining the Alignment (diagCM)
bull ELAND is part of the standard pipeline
bull ELAND accurately places reads in the correct genomic location
bull Use cross-match a Smith-Waterman aligner to improve local alignment
12
Eland version XXX
ELAND Aligned Reads
eland_ms_32 Casava 17
Cross-match Improved Alignment
13
Comparison of Aligners (simulated exome data)
Perc
enta
ge C
orre
ct
85
90
9510
0
-6 -4 -2 0 2 4 6
Variant Size
BayesianGenotypeCalling MostProbableGenotype
ACCTTCCTCGAGGCTAGCCTAGCCGGGGTGGGCAAGGCCTTCCGGGGGGGTAACCTCCTCCAAAACCTCCCTATGCCCCCCCAATTTTTTATATATATCCTCCTCCAA A
A
A A
A
A
A A
A A
A A
P(Genotype|Data)
AT 0 AA -14
Difference between most probable AA 10-6
10-15 and next most probable genotype AC 10-15AG AC -34
AT 099999981 AG -34 10-75CC TT -124 MPG Score = 14 10-75 CT -138CG 10-60CT Convert to natural Log GT -138
GG 10-75 Probabilities and Sort CC -172 10-60GT CG -172 10-54TT GG -172
14
MPG of Haploid Regions
bull Human autosomes are normally diploid bull MPG is designed to call two alleles for the
autosomes and X chromosome if the sample is from a female
bull For samples from males MPG is run in haploid mode on the non-sudoautosomalregions of X and Y
bull Thus only testing for the four nucleotides
Exome Coverage versus Input Sequence
15
Num
ber o
f Sam
ples
SureSelect 38Mb
SureSelect 50Mb
Illumina TruSeq 62Mb
0 20
0 40
0 60
0 80
0 10
00
Whole Exomes Processed at NISC
Mar 2010 Dec 2010 Sep 2011
NBEAL2
16
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
Targeted Regions Depth-of-Coverage Histogram
0
0005
001
0015
002
0 50 100 150 200
Depth of Coverage
Frac
tion
per B
in
TruSeq_example Poisson
Refining the Alignment (diagCM)
bull ELAND is part of the standard pipeline
bull ELAND accurately places reads in the correct genomic location
bull Use cross-match a Smith-Waterman aligner to improve local alignment
12
Eland version XXX
ELAND Aligned Reads
eland_ms_32 Casava 17
Cross-match Improved Alignment
13
Comparison of Aligners (simulated exome data)
Perc
enta
ge C
orre
ct
85
90
9510
0
-6 -4 -2 0 2 4 6
Variant Size
BayesianGenotypeCalling MostProbableGenotype
ACCTTCCTCGAGGCTAGCCTAGCCGGGGTGGGCAAGGCCTTCCGGGGGGGTAACCTCCTCCAAAACCTCCCTATGCCCCCCCAATTTTTTATATATATCCTCCTCCAA A
A
A A
A
A
A A
A A
A A
P(Genotype|Data)
AT 0 AA -14
Difference between most probable AA 10-6
10-15 and next most probable genotype AC 10-15AG AC -34
AT 099999981 AG -34 10-75CC TT -124 MPG Score = 14 10-75 CT -138CG 10-60CT Convert to natural Log GT -138
GG 10-75 Probabilities and Sort CC -172 10-60GT CG -172 10-54TT GG -172
14
MPG of Haploid Regions
bull Human autosomes are normally diploid bull MPG is designed to call two alleles for the
autosomes and X chromosome if the sample is from a female
bull For samples from males MPG is run in haploid mode on the non-sudoautosomalregions of X and Y
bull Thus only testing for the four nucleotides
Exome Coverage versus Input Sequence
15
Num
ber o
f Sam
ples
SureSelect 38Mb
SureSelect 50Mb
Illumina TruSeq 62Mb
0 20
0 40
0 60
0 80
0 10
00
Whole Exomes Processed at NISC
Mar 2010 Dec 2010 Sep 2011
NBEAL2
16
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
Eland version XXX
ELAND Aligned Reads
eland_ms_32 Casava 17
Cross-match Improved Alignment
13
Comparison of Aligners (simulated exome data)
Perc
enta
ge C
orre
ct
85
90
9510
0
-6 -4 -2 0 2 4 6
Variant Size
BayesianGenotypeCalling MostProbableGenotype
ACCTTCCTCGAGGCTAGCCTAGCCGGGGTGGGCAAGGCCTTCCGGGGGGGTAACCTCCTCCAAAACCTCCCTATGCCCCCCCAATTTTTTATATATATCCTCCTCCAA A
A
A A
A
A
A A
A A
A A
P(Genotype|Data)
AT 0 AA -14
Difference between most probable AA 10-6
10-15 and next most probable genotype AC 10-15AG AC -34
AT 099999981 AG -34 10-75CC TT -124 MPG Score = 14 10-75 CT -138CG 10-60CT Convert to natural Log GT -138
GG 10-75 Probabilities and Sort CC -172 10-60GT CG -172 10-54TT GG -172
14
MPG of Haploid Regions
bull Human autosomes are normally diploid bull MPG is designed to call two alleles for the
autosomes and X chromosome if the sample is from a female
bull For samples from males MPG is run in haploid mode on the non-sudoautosomalregions of X and Y
bull Thus only testing for the four nucleotides
Exome Coverage versus Input Sequence
15
Num
ber o
f Sam
ples
SureSelect 38Mb
SureSelect 50Mb
Illumina TruSeq 62Mb
0 20
0 40
0 60
0 80
0 10
00
Whole Exomes Processed at NISC
Mar 2010 Dec 2010 Sep 2011
NBEAL2
16
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
Comparison of Aligners (simulated exome data)
Perc
enta
ge C
orre
ct
85
90
9510
0
-6 -4 -2 0 2 4 6
Variant Size
BayesianGenotypeCalling MostProbableGenotype
ACCTTCCTCGAGGCTAGCCTAGCCGGGGTGGGCAAGGCCTTCCGGGGGGGTAACCTCCTCCAAAACCTCCCTATGCCCCCCCAATTTTTTATATATATCCTCCTCCAA A
A
A A
A
A
A A
A A
A A
P(Genotype|Data)
AT 0 AA -14
Difference between most probable AA 10-6
10-15 and next most probable genotype AC 10-15AG AC -34
AT 099999981 AG -34 10-75CC TT -124 MPG Score = 14 10-75 CT -138CG 10-60CT Convert to natural Log GT -138
GG 10-75 Probabilities and Sort CC -172 10-60GT CG -172 10-54TT GG -172
14
MPG of Haploid Regions
bull Human autosomes are normally diploid bull MPG is designed to call two alleles for the
autosomes and X chromosome if the sample is from a female
bull For samples from males MPG is run in haploid mode on the non-sudoautosomalregions of X and Y
bull Thus only testing for the four nucleotides
Exome Coverage versus Input Sequence
15
Num
ber o
f Sam
ples
SureSelect 38Mb
SureSelect 50Mb
Illumina TruSeq 62Mb
0 20
0 40
0 60
0 80
0 10
00
Whole Exomes Processed at NISC
Mar 2010 Dec 2010 Sep 2011
NBEAL2
16
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
MPG of Haploid Regions
bull Human autosomes are normally diploid bull MPG is designed to call two alleles for the
autosomes and X chromosome if the sample is from a female
bull For samples from males MPG is run in haploid mode on the non-sudoautosomalregions of X and Y
bull Thus only testing for the four nucleotides
Exome Coverage versus Input Sequence
15
Num
ber o
f Sam
ples
SureSelect 38Mb
SureSelect 50Mb
Illumina TruSeq 62Mb
0 20
0 40
0 60
0 80
0 10
00
Whole Exomes Processed at NISC
Mar 2010 Dec 2010 Sep 2011
NBEAL2
16
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
Num
ber o
f Sam
ples
SureSelect 38Mb
SureSelect 50Mb
Illumina TruSeq 62Mb
0 20
0 40
0 60
0 80
0 10
00
Whole Exomes Processed at NISC
Mar 2010 Dec 2010 Sep 2011
NBEAL2
16
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
NBEAL2 5rsquo
Exome Variation Statistics TruSeq 62Mb Male Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 133047403 142361 000072
Auto 125491045 139295 000076
chrX 6842299 2600 NA
chrY 710243 1435 NA
17
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
Exome Variation Statistics TruSeq 62Mb Female Sample
Type
Total Genotype
Calls SNVs Within-sample Heterozygosity
Total 125681915 136993 000075
Auto 120559746 132616 000076
chrX 5096376 2701 000034
Example Heterozygous SNV
18
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
Example Heterozygous Deletion
Coverage of MPG gt= 10 Genotype Calls
Total Raw Sequence
Aligned Sequence
Genotype calls CCDS
Genotype calls UCSC coding
SureSelect 38Mb 67 Gb
50 Gb (131x)
89 74
SureSelect 50Mb 105 Gb
61 Gb (122x)
89 85
TruSeq 62Mb 90 Gb
71 Gb (114x)
91 89
Whole Genome Shotgun
192 Gb 133 Gb (44x)
86 83
19
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
GenotypeConcordance
Total Agreement with Genotype Chip (CCDS)
Whole Genome Shotgun 99908
SureSelect 38Mb 99910
SureSelect 50Mb 99857
TruSeq 62Mb 99865
Whole Exome Sequencing bull Being applied to
ndash Undiagnosed Diseases Program (100rsquos of samples)
ndash ClinSeq (gt1000 samples) ndash Variety of other PI driven projects (eg cancer)
bull Data generation rate per year ndash 200 exomes per GAiiX ndash 1200 exomes per HiSeq 2000
bull Analysis results ndash Genotype data for 90 of consensus coding exon
bases (CCDS) ndash Accuracy of genotype calls over 995
20
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
Exome Sequencing Pipeline
Sample DNA Fragmentation
Illumina Library Preparation
Exome Enrichment
Cluster Generation
Sequencing and Basecalling
Sequence Read Alignment
Variation Detection
Variant Annotation and Working With Whole-Exome Data
bull One sample produces gt 100k variants
bull One hundred samples gives rise to 600k or more
bull How does one work with such large datasets
bull The next speaker Dr Jamie Teer will address these next steps
21
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
Acknowledgements NIH Intramural Sequencing bull Mullikin Lab
Center ndash Nancy Hansen
bull Sequencing Operations ndash Pedro Cruz ndash Bob Blakesley ndash Praveen Cherukuri ndash Alice Young ndash Lab Staff bull Biesecker Lab
bull Bioinformatics ndash Jamie Teer ndash Gerry Bouffard ndash Baishali Maskeri ndash Jenny McDowell ndash Meg Vemulapalli
bull IT Linux Support ndash Jesse Becker ndash Matt Lesko
httpresearchnhgrinihgov
22
23
23
